WO2001065789A1

WO2001065789A1 - Method and device for estimating channel propagation

Info

Publication number: WO2001065789A1
Application number: PCT/FR2001/000542
Authority: WO
Inventors: Lionel Hayoun; Didier Pirez
Original assignee: Thales
Priority date: 2000-02-29
Filing date: 2001-02-23
Publication date: 2001-09-07
Also published as: US20030053522A1; US7103093B2; FR2805689B1; DE60132561D1; EP1260071B1; FR2805689A1; AU2001237489A1; EP1260071A1; ATE385112T1

Abstract

The invention concerns a method for estimating a propagation channel, between a data transmission source and a receiver, using at least a learning sequence comprising so-called pilot symbols known to the receiver, to decode a data signal (digital signal), the signals being transmitted by successive frames, each frame being partitioned into a number of time slots. The method comprises at least the three following steps: a) selecting a search window, estimating the complex amplitude and energy of each predominant path k located in the search window during each one of the time slots and demodulating the data bits (useful bits); b) updating the energy of each possible path of the search window all the K slots; c) selecting all the K*L slots, the predominant energy paths k; applying to the demodulation a CDMA-type, or a UMTS-type signal.

Description

METHOD AND DEVICE FOR ESTIMATING A SPREAD CHANNEL

The present invention relates to a method and a device for estimating the parameters of a propagation channel, for decoding digital signals, using a training or reference sequence known to the receiver. The signals are transmitted in successive frames, each frame being partitioned into a determined number of time slots.

The invention applies, for example, to demodulate a wideband or spread spectrum digital signal. The invention finds its application in the field of cellular networks operating on the principle of "CDMA" (English abbreviation for Code Division Multiple Access), such as networks based on the IS95 standard in service for several years in the USA, or even those based on the UTMS standard (English abbreviation for Universal Mobile Telecommunication System) which should be operational in Europe at least by the year 2002. It applies in particular within the framework of mobile radiocommunications.

The method according to the invention can without departing from the scope of the invention apply to propagation channels of any kind.

In a CDMA system all users share the same band and use spectrally spread waveforms using codes, each user having their own code.

As a guide, a description of such a system can be found in one of the following references: [2] W.C.Y. LEE, "Overview of cellular CDMA", IEEE Transactions on vehicular technology, vol. 40, n ° 2, May 1991 or

[3] RL PICKHOLTZ, DL SHILLING and LB. MILSTEIN, "Theory of spread- spectrum communications - A tutorial", IEEE Transactions on communications, vol. com-30, n ° 5, pp 855-884, May 1982. The principle of spread spectrum by direct sequence is to cut a useful bit of duration T into N _G pulses or chips of duration T'≈T / NG . The band W of the resulting signal is thus N _G times greater than that of the useful signal W, which theoretically allows a modulation rhythm N _G times greater. The sequences of pulses or codes, are generally pseudo-random sequences (PN sequences), designed to have specific autocorrelation and intercorrelation properties (ideally, autocorrelation is impulse and the correlations between codes are zero). Such sequences are described for example in the two works: [1]: JG PROAKIS, "Digital communications", McGraw Hill and the reference [2] mentioned above.

Upon reception, each user (or transmission channel) is demodulated by a correlation operation with its spreading code, known to the receiver. This processing is a suitable filtering which provides a reception gain equal to the ratio between the spread band and the initial useful band shown in FIG. 1.

The CDMA concept allows not only multiple access (large combinatorial offered by pseudo-random codes, to assign a different code to each user) but also increased robustness to interference (intentional or not) and poor propagation conditions (paths multiple, fading, ...).

Currently, it is known to use a spectrally spread waveform to effectively resist the phenomenon of multipath (this phenomenon appears for various propagation channels and in particular for the radiomobile channel) and even to take advantage of it to combat fading or fading (random variations in the amplitude and phase of the signal).

Indeed, if T _m is the temporal spreading of the channel and if W is the bandwidth of the transmitted signal, then it is theoretically possible to isolate in reception T _m W components (a bandwidth W allows in reception, by correlation, to isolate paths with a time resolution equal to 1 / W), which physically correspond to the different propagation paths. All or part of these components are independent and it is possible to recombine them so as to increase or maximize the signal to noise ratio. A diversity effect is thus obtained exploiting the statistical independence of the different propagation paths constituting the received signal. This diversity technique was initially proposed by Price and Green in 1958, who gave the name of RAKE receptor to optimal treatment (the receptor isolates each path and then makes a recombination, hence the analogy with a garden rake). This processing, which coherently sums all the paths, simply consists of a filtering adapted to the received signal, that is to say adapted to the channel (we show that this processing is optimal when we consider the reception of a single user or code, all the other users constituting a noise having the property of being white and Gaussian).

In this regard, one will refer for example to the work "Digital communications" with the editions McGraw Hill having for author J.G. Proakis.

The implementation of the RAKE receiver therefore supposes the knowledge of the impulse response of the channel, that is to say the complex amplitudes of all the paths. In practice, it is possible to have an estimate of these amplitudes which takes into account the different characteristics of the propagation channel (time spread, fading) and in particular its different fluctuation scales (fast fading and slow fading).

On the radiomobile channel, there are generally two scales of variation of the propagation characteristics described for example in the work having for author D. Parons published by Pentech Press in 1992 under the title "The mobile radio propagation channel".

The first scale corresponds to fast fading (or Rayleigh fading), which induces a variation of the amplitudes and phases of the paths after a time of the order of the coherence time of the channel (inverse of the Doppler band of the channel). In the mobile radio context, this decorrelation time (time interval for which two realizations of the complex amplitude of a path are decorrelated) corresponds to an advance of the mobile of the order of half the wavelength of the frequency reference carrier [4]. On this time scale, the delay times of the journeys are constant.

The second fluctuation scale corresponds to slow fading (or long-term fading), due to masking effects and changes in the environment. This phenomenon induces on the one hand a modification of the average power of the paths, and on the other hand a significant evolution of the delay times of the paths; journeys are also likely to appear and others to disappear.

The decorrelation time (or decorrelation distance) associated with slow fading depends on the channel, that is to say on the type of environment in which the mobile operates. In practice, this decorrelation time is much higher (at least an order of magnitude) than that associated with fast fading. For example, the ETSI within the framework of UMTS, indicates a distance of decorrelation of slow fading, for an urban environment, approximately equal to 20 meters; this leads to a report approximately 300 (20 / (λ 2) with λ = c / f and f = 2.10 ⁹ Hz) between the long and short term decorrelation times.

The object of the invention relates to a method allowing the dynamic estimation of the propagation channel, taking into account the various fluctuations mentioned previously, with the aim of demodulating a digital signal. The process applies in particular in the following cases:

• Single-user reception, ie we are interested in the demodulation of a single user, all the others being considered as noise (no joint detection or multi-user detection);

• Transmission of a reference sequence, in addition to the useful data, dedicated to the estimation of the propagation channel;

• The propagation channel is fluctuating;

• The time spread of the channel is less than the duration of a useful bit. Thus, the inter-symbol interference can be neglected and the implementation of an equalizer downstream from the RAKE receiver is not necessary. Each useful bit or symbol is demodulated independently of the other useful bits or symbols;

• The channel can be considered constant for the duration of a useful bit or symbol (T _CO erence »T, where T _coherence is the coherence time of the channel).

The word "symbol" designates an information unit, the expression "learning symbol" or "pilot" corresponds to an information unit of the known signal.

The invention relates to a method for estimating a propagation channel, between a data transmission source and a receiver, using at least one training sequence comprising symbols known as “pilot symbols”, said sequence being known. of the receiver, to decode a data signal (digital signal), the signals being transmitted in successive frames, each frame being partitioned into a determined number of time slots. It is characterized in that it comprises at least the following steps: a) choosing a search window, estimating the complex amplitude and the energy of each predominant path k located in the search window during the duration of each of the time slots and demodulating the data bits (useful bits), b) updating the energy of each possible path of the search window every K slots, and c) selecting all the K ^* L slots, the preponderant energy paths k.

The invention also relates to a device making it possible to estimate a propagation channel between a transmission means and a reception means in order to demodulate the useful data received. It is characterized in that it comprises at least one microprocessor or device suitable for implementing the method according to the invention, the signal possibly being a spread signal and being demodulated according to the RAKE principle.

The device and the method according to the invention apply to demodulate a spread band signal of CDMA type, such as a signal of UMTS type.

The method according to the invention notably offers the advantage of having one or more estimated parameters of a propagation channel "in real time", of updating these parameters in order to improve the demodulation of the signals received by a receiver operating on the RAKE principle. The improvement consists in particular in overcoming variations in the parameters of the channel such as time spread, variation in amplitude and in phase of a path.

The invention will be better understood on reading the description which follows by way of illustration and in no way limitative of an example relating to the estimation of a digital signal propagation channel of the UMTS type where: • FIG. 1 represents a diagram spectrum spreading,

FIG. 2 diagrams a waveform for the uplink of a transmitter to a receiver,

FIG. 3 diagrammatically shows a synoptic transmission-reception of a UTMS wave according to the invention, FIG. 4 diagrammatically shows the periods of realization of the three stages of the method according to the invention on the frame of a signal,

FIG. 5 is an example of a flow diagram of an estimation of a propagation channel implemented in the mode of the invention more particularly described below, and • Figures 6 to 9 detail the steps of Figure 5 in the form of flowcharts.

Before setting out the different stages of the process according to the invention, the aim of the following paragraph is to allow the reader to understand the principles of the RAKE receiver used in the example given by way of illustration and in no way limiting.

Consider the emission of a signal modulated in BPSK (Binary Phase Shift

Keying) at rhythm T _s (extension to other phase modulations does not pose any problem: the receiver structure is identical, only the downstream decision procedure is modified) and spread by direct sequence with a gain N = T _s / T _c where T _c is the duration of the chip. The transmitted signal is therefore of the form (baseband): st ≈Vξ ^' cK cα)

where P _s is the signal strength, and d (t) the sequence of useful bits, d (t) = Σ d _k g _τ (t - kT _s ) k = -

d = ± 1, equiprobably, g _Ts is the pulse used to convey the information, assumed to be rectangular here for the sake of simplicity:

g _Tt (t) = l if O ≤ t ≤ T, g _τ (t) = 0 otherwise c (t) is the spreading code ideally having a pulse autocorrelation function (that is to say that the correlation between two versions of the code shifted by any nonzero time is 0),

with c _k = ± 1

If the channel impulse response has the expression:

h (t) = ∑ a _k e ^jφk δ (t * - τ _k ) k = l

that is to say that the propagation channel consists of N, paths, of complex amplitudes a _k e ^{) ok} and of propagation time τ _k , with Vi ≠> T _C

(in other words all the paths are discernible) then the received signal (baseband complex) has the expression:

where n (t) represents the thermal noise, assumed centered Gaussian complex, temporally white and decorrelated from the useful signal, of spectral density N ₀ .

The principle of RAKE is to isolate the contributions of each path and then recombine them in phase coherence so as to optimize the signal to noise ratio.

In the following, it will be assumed that the inter-symbol interference is negligible, in other words that the duration of a symbol is large compared to the time spread of the channel. This assumption is often verified in practice and is adopted, explicitly or implicitly, by most authors. The RAKE receiver will therefore be implemented independently on each symbol d _k .

It is also assumed that the channel is constant for the duration of a useful symbol (Δt _c »T _s , Δt _c being the coherence time of the channel).

The contribution of a path is obtained by correlating the received signal with the spreading code offset by the delay of the path considered, assumed to be known.

Thus, the contribution of the delay path τ, calculated on the symbol of index k, is given by

(for the sake of simplicity, we assume that the signal samples are received at the chip rate):

The development of (E2) using (E1) leads to three terms, the first constituting the useful part and the other two constituting the parasitic part:

The spreading code is designed so that two offset versions, of a time greater than the duration of the chip, are approximately decorrelated (inter-path interference is thus minimized): χ _s J ^j ₍ ( _Ts ) c ( ^•• t - τ, i) 'c ( ^{• »} t - X _j ) dt ≈ - _N « 1 pourlτ. - τJ≥T ,. (P,)

The contribution of the path i can therefore be written: r _I (k) = / p7a ₁ e ^jφ - d _k + b ₁ (k) b, (k) being the parasitic term grouping the thermal noise and the interference created by the other paths. A simple calculation shows that in the absence of inter-path interference, the noise power b, (k) is No / T _s .

The problem is now posed as follows: having the N _t contributions r, (k) (we assume here that the number of taps on the receiver is equal to the real number of paths), we are looking for a set of complex weights (w * ,, w ₂ , ..., w _Nt ) such that the signal to noise ratio of the recombined signal r (k) is maximum:

r (k) = ∑ ₁ * r, (k)

where w ^* is the conjugate complex of w ,.

The signal to noise ratio of r (k) is defined by (E (x) denotes the average value of x):

If we assume that the noise terms associated with each path b, (k) are decorrelated and have the same mean power P then the optimal weighting is given by (Schwartz inequality): w _{ι =} ka, e ^jφi

where k is an arbitrary constant.

Each path is therefore weighted in accordance with its own signal-to-noise ratio. Ultimately, the optimal RAKE receiver therefore calculates, for the symbol k, the quantity (assuming perfect the estimation of the channel parameters):

+ b (k) (E3)

N, with b (k) = X a ₁ e- ^{, φl} b, (k) ι = I

and the symbol is decided by comparing the real part (theoretically, except for noise, r (k) is real) from r (k) to 0 (in BPSK).

The signal to noise ratio obtained is:

RSB _RAKE .

It is the sum of the signal-to-noise ratios relating to each of the paths.

RAKE processing is optimal in the absence of interference between symbols (in the strict sense, this means that the propagation channel contains only one path or that the autocorrelation function of the code is ideal) and if the noise ( thermal noise + inter-user interference) is white (which means that the codes of the different users are orthogonal for all possible time offsets).

The implementation of the RAKE receiver, described by equation (E3), assumes that the complex amplitudes a _k e ^iok and the delays τ _k of the paths are known. In practice, only an estimation of these parameters can be obtained, which in general is carried out by means of the emission of a sequence of symbols known to the receiver (designated by the term reference sequence or pilot sequence or alternatively sequence of learning). The channel estimate must be updated at a frequency consistent with the variation speeds (Doppler spreading) of the different parameters a _k , φ _k and τ _k . Finally, if the channel varies slowly with respect to the duration of a symbol (T _s ), it is possible to make this estimate using several consecutive symbols which makes it possible to obtain more reliable estimates.

The example of implementation is described below in an application of cellular networks operating on the principle of CDMA (Code division Multiple Access), as mentioned above.

FIG. 2 describes a UTMS type waveform used for the uplink, that is to say from a mobile station to a base station. The structure of the wave frame is composed for example of several slots Si or time slots, i being the index of the slot.

Each slot comprises the useful data corresponding to the DPDCH slot (abbreviated to English Dedicated Physical Data Channel), for example the Ndata number and the control data or DPCCH slot (abbreviated to Dedicated Physical Control Channel) which are transmitted in parallel via a complex multiplexing on phase and quadrature channels. The useful data are transmitted on the "in phase" channel (channel I) (FIG. 3) and the control data are transmitted on the "quadrature" channel (Q channel) on this same figure 3. The useful and control data are modulated for example in BPSK

(Binary Phase Shift Keying) and the signal obtained after multiplexing is therefore modulated in QPSK (Quadrature Phase Shift Keying).

Each frame, of duration 10 ms, consists for example of 16 time slots of duration 0.625 ms. The control sequence shown schematically in Figure 2 is composed of

Npilot bits known to the receiver used in the method of channel estimation according to the invention, control bits used for power control Ntpc bits, and bits indicating the bit rate used on the payload channel Nri bits. The total number of control bits is for example fixed at 10. The digital distribution and the arrangement of these three types of control will be fixed for example by the UMTS standard exposed in the reference [5] ETSI, “Universal Mobile Telecommunications System ( UMTS); UMTS Terrestrial Radio Access (UTRA); Concept evaluation ", UMTS 30.06 version 3.0.0, TR 101 146 V3.0.0 (1997-12).

The useful and control data bits are for example spread by orthogonal Walsh codes [5], so as to join the chip or pulse rate at 4.096 MHz.

The spreading factor (or spreading gain) used for the useful bits is chosen for example as a function of their bit rate, the lower the useful input bit rate, the lower this spreading factor.

By designating by k the number of bits for the useful data (DPDCH) and control (DPCCH) channels, it is linked to the spreading factor SF by the relation SF = 256/2 ^k . The spreading factor can therefore take values between 256 and 4. The spreading factors relating to the data channel and to the control channel can be different insofar as the number of bits or symbols carried by slot may be different.

FIG. 3 represents a diagram of spreading, modulation and scrambling of a signal. After spreading by their respective code, the signal is scrambled by a primary scrambling code linked to the mobile station (C ' _scramb ) - The resulting signal can, optionally, be scrambled a second time by a secondary scrambling code (C'scramb) -

The primary scrambling code is a complex code generated from KASAMI (ref) sequences of length 256. The secondary scrambling code is 40,960 chips long (10 ms duration) and constitutes a segment of a code of Gold (ref) of length 2 ⁴¹ - 1. The signal transmitted from the mobile station is formatted in the block

10, it is then transmitted to a transmission filter 11, then it is received at the base station. This base or reception station comprises a reception filter 13 followed by a reception block 14 comprising for example a receiver 15 of the RAKE type and means such as a microprocessor 16 programmed for implementing the steps of the method according to the invention, in order to estimate the parameters of the propagation channel and demodulate the digital signals.

Description of the steps implemented during the process

The principle of the invention is to demodulate the useful bits or symbols using a demodulation algorithm consisting of two parts, the first consists in estimating the channel and the second in demodulating the useful symbols or bits. The characteristics of the propagation channels such as time spread, Doppler spread can be arbitrary.

The second part is carried out according to the RAKE algorithm known to those skilled in the art, details can be found in the aforementioned Proakis reference.

According to the invention, the estimation of the channel is carried out by implementing an algorithm comprising at least the three steps described below.

These steps are carried out for example by processing the digital signal received after filtering adapted to reception by the function of shaping the pulses of Nyquist root type for example and correcting a possible Doppler shift.

First stage :

All time slots, every Δt-i seconds, for example every 0.62 ms, the process:

* estimates the complex amplitude of all the selected paths, by correlation with the reference sequence. The selected paths correspond to the paths selected during the third step or for the initialization of selected paths by comparing the energy of a path with a threshold value, * "demodulates the useful bits by phase recombination of these paths.

Second step

Every K slots, every Δt ₂ seconds, the method updates the energy of the paths whose delays belong to the search window with a given analysis step. This step combined with the third step allows long-term monitoring of the paths present in the window.

Third step :

Every K ^* L slots, every Δt ₃ seconds, the process updates the selection of the predominant routes taken into account for the RAKE demodulation. The periods of realization Δt-, Δt ₂ , Δt ₃ of these steps expressed in seconds are for example chosen to verify the relation Δt-> Δt ₂ > Δt ₃ .

In the following description, • Certain steps of the algorithm, known to those skilled in the art, are not detailed (Nyquist filtering on reception, management of the data flow, etc.);

• It is assumed that the correction of a possible Doppler shift is carried out;

• Vectors and matrices are noted in bold;

• X ^' is the conjugate complex of the scalar or the vector X; x ^{1 *} y is the scalar product (Euclidean or Hermitian) of the vectors x and y; the vector x described between index_start and index_end with a resol step is noted: x (index_start: index_end: resol) At the input of the algorithm, the sample file is composed of complex samples obtained after half-Nyquist filtering ( transmission) and oversampling of the UMTS signal.

At the output of the algorithm, the file obtained is composed of a file of demodulated and despread data bits (complex values). These bits are ready to be decided, for example according to the principle stated above.

The list of key parameters of the invention and of the variables used in the algorithmic description is given in the form of two tables 1 and 2 appended.

The process takes place according to steps 21 to 27 of the flow diagram shown in FIG. 5. These steps are applied for each time slot, that is to say every Δti seconds. The method begins with an initialization step 20 of different values, in particular: the size of the search window expressed in seconds and in samples: Taille_fen_rech, Etal_temp_max; the delay sub-sampling factor Sous_ech_retard; the forget factor λ used during the second step, the threshold value Threshold_power for the selection of journeys during the third step ; the frequency of updating the powers K and the factor for updating the delays L.

It also includes the initialization of a table comprising the predominant paths determined for example by comparing the energy of a path with the threshold value and by retaining only the paths whose energy is greater than this threshold value.

The samples are filtered using a half-Nyquist type filter, for example.

The method begins in step 21 by calculating the complex amplitude of the paths selected during the third step. The complex amplitudes are obtained for example by correlation between the received pilot symbols (which are spread) and the code relating to the control bits (product of the spreading code and of the primary scrambling code), the elimination of the contribution of the pilot bits and averaging (by simple summation of the estimated amplitudes obtained with each of the symbols) on the pilot N_symbol which in particular makes it possible to increase the signal / noise ratio.

The correlations carried out take into account, at the signal level of the pilot symbols received, the delay times identified during step 3.

The estimation of ^the amplitude of the paths retained for complex demodulation for each pilot symbol is performed by the following sequence of steps described in relation with Figure 6

Step 30

For j = 1 at Nb_symbol_pilot

Positioning at the beginning of the pilot symbol number j Ind_debut_demod = Ind_centre_fen + (j - 1) * GE_contrôle * Nb_ech_chip

Step 31 -Calculation of the correlation for all the paths selected For k = 1 to Nb xajects

CorreIation_pilote (k, j) = Code_etaI_control * SignaI (Ind_debut_demod

+ Delays (k) * Sous_ech_retard: Ind_debut_demod

+ Delays (k) * Sous_ech_retard + (GE_control - 1) * Nb_ech_chip: Nb_ech_chip) * Seq_symboI_pilot (j) ^* End "For k = 1 at Nbjrajets"

End "For d = 1 at Nb_symbol_pilot"

Step 32 Calculate the ^amplitude of the complex channel

For k = 1 at Nbjxajets

Channel _ estimated (k) =

pilot {k, j)

Nb _ sym o s _ pilot ^ f

End "For k = 1 at Nbjxajets"

The following step 22 (FIG. 5) is a step of demodulating useful symbols comprising a step of despreading and demodulation according to the RAKE principle, and the sequence of steps described below and in FIG. 7:

For j = 1 at Nb_symbol_data

Step 33 - Positioning at the beginning of the useful symbol number j Ind_debut_demod = Ind_centre J ^~ enetre + (j - 1) * GE_data * Nb_ech_chip

Step 34 - Despreading

For k = 1 to Nb discharges

CorreIation_data (kj) = Code_etal_data ¹ * Signal (Ind_debut_demod +

Delays (k) * Sous_ech_retard: Ind_debut_demod + Delays (k) * Sous_ech_retard + (GE_control - 1) * Nb_ech_chip: Nb_ech_chip)

End "For k = 1 at Nbjrajets"

Step 35 - Calculation of useful symbols demodulated by RAKE recombination

Symbol_data_demod (j) = Estimated_Channel (1: Nbjrajets) ¹ * Correlation lata (l: Nbjrajets, j)

End "For j = 1 at Nb_symbol_data The second step of the process carried out every K slots, every Δt ₂ seconds and corresponding to step 23 (Figure 5) consists in updating the energy of each path located in the search window and determining the complex amplitude of the journey. These two steps are detailed in FIG. 8. The method performs (step 37) the correlation between each pilot symbol received and the control code, for all the possible delays included in the search window and with a given resolution. The resolution varies for example between the duration of the sample and the duration of the chip.

The search window is for example centered on the current center of the impulse response of the channel, the center being determined for example in step 3, and its size being equal to twice the maximum time spread taken into account by the algorithm.

In a second step (step 38), the method averages the complex amplitudes estimated at the level of the pilot symbol, over all of the pilot symbols, so as to increase the quality of the estimation, this for all the possible paths.

Finally, the method (step 39) calculates the instantaneous power C (k) of these paths (square of the module of complex amplitudes) which is used to obtain the average power C | _0n g_ _term e (k), using the following formula (first-order low-pass filtering):

C | oπg_terme (k) = λ C | _on g_teπ * ne (k) + C (k)

where λ is a forgetting factor whose value, like that of the parameter K, must be chosen according to the characteristics of the propagation channel.

This smoothing overcomes rapid fading, which can occasionally cause a drop in the instantaneous power of a strong trip.

Update of the energy of each path contained in the search window

This update begins with the determination of a table, step 37 containing, for all the possible delay times located in the search window and for each pilot symbol, the correlations between the pilot symbols received and the spreading sequence of control lo

Pourj = 1 at Nb_symbole_pilot

Positioning on the pilot symbol number j

Ind_debut_demod = Ind_centre enetre + (d-1) * GE_contrôle ^x

Nb_ech_chip

Correlation between the pilot symbols received and the control sequence

For k = 1 to Jen_rech size

Correlation_pilote (kj) = Code_etal_control ^±

* Signal (Ind_debut_demod + k * Sous_ech_retard: Ind_debut_demod + k * Sous_ech_retard + (GE_control - 1) * Nb_ech_chip: Nb_ech_chip) * Seq_symbole_pilot (j) ^*

End "For k = 1 to Taillejfen_rech" End "Pourj = 1 to Nb_symbole_pilot"

The next step 38 consists for each journey in the search window in averaging the correlations on the Nb_symboles_pilot, in calculating the modulus of the correlation and in updating the average energy of the journey.

For k = 1 at SizeJ ^" en_rech

Correl _ pilot _ avg (k) = pilot (k, j)

Puiss_inst_trajet (k) = module (Correl_pilot_moy (k))

Power kerjtrajet (k) = λ * Power_moy: rajet (k) + Power nstjirajet (k) ²

End "For k = 1 at Size J ^" en_rech "

The following step 39 corresponds to the calculation of the complex amplitude of the channel

For p = 1 to N jrajets

Estimated_Channel (p) = Correl_pilot_moy (Delays (p))

End "For p = 1 at Nbjrajets" Step 3 or third step of the process is carried out every K ^* L slots, every Δt ₃ seconds for example by scrolling detailed in FIG. 9 composing step 24 for updating the predominant routes which are used during the first step of the method for demodulating the received signals. The progress of these steps begins, for example, as follows: a step 50 for calculating the maximum average power, followed by a step 51 for selecting all the paths for which the ratio between their average power and the average power exceeds one threshold, for example 0.5.

According to a variant implementation of the method, after the selection of the paths (step 51), the method continues with a step 52 of choosing the best sampling comb, the paths retained in fine are separated by an integer number of chips.

From the selected paths, the method then determines, for example then using steps and methods known to those skilled in the art, the number of paths selected on the best comb and the vector Delays of the indices of the paths selected, step 53 .

Step 54 consists in calculating the complex amplitude of the selected paths.

The following step 55 comprises an update of the center of the search window, of tables containing the values of the delays and of the average powers for the selected paths.

Update of the choice of routes retained for demodulation

Step 50 - Calculation of the maximum average power (among all possible routes)

Power noy_max = MAX (Puiss_moy_trajet)

Step 51 - Selection of journeys whose average power exceeds the threshold

For k = 1 at Size J ^~ en_rech

IF (Puiss_moyJτajet (k)> Puiss_moy_max * Seuil_puissance) Then Selec Jxajets (k) = Puiss_moy Jxajet (k)

Otherwise Selec jtrajets (k) = 0 END IF End "For k = 1 at Taillejenjech"

Step 52 - Determination of the best sampling comb Firstly, calculate the cumulative average power over all the paths exceeding the threshold and located on the same sampling comb (the paths of a given comb are separated by a number whole crisps)

Nb_retard_chip = Nb_ech_chip / Sous_ech_retard For k = 1 at Nb_retard_chip - 1

Size _ fen_rech

Power _ tot _ comb (k) = ∑ Selec _ paths (k + p * Nb_ delay_chip) p = 0

End "For k = 1 at Nb_retard_chip - 1"

Then in a second step the choice of the best sampling comb by determining the maximum total cumulative power:

Ind_meilleur_peigne = ARG (MAX (Puissjtot_peigne))

Step 53 - from the selected routes, the method calculates using steps known to those skilled in the art

• the number Nb of journey paths selected on the best comb, which it can compare to a determined number for example Nb_trajets_max, and • the Delays vector (l: Nb_trajets) of the indices of the paths retained on the best comb.

The method then calculates during step 54 and for the paths selected in step 52 The complex amplitude or the estimate of the current channel

For p = 1 at Nbjrajets

Estimated_Channel (p) = Correl_piIot_moy (Delays (p))

End "For p = 1 at Nbjrajets"

Step 55 includes updating the center of the impulse response of the current channel so as to update the position of the center of the search window by calculating the average delay:

Average delay = 0.5 * (Delaysd) + Delays (Nbjrajects)) And the update of the Delays and Power tables, as well as the update of the index of the center of the search window lnd_centre_fenetre

Delays (l: Nbjrajets) = Delays (l: Nbjrajets) - Average_delay Ind_centreJ ^~ enetre = Ind_centreJenetre + Delard_moyen * Under delay

In the case where the Num_slot parameter is not a multiple of K * L, the method processes the time slot or slot using a method known to those skilled in the art to determine the amplitude and the phase of each of the journeys during step 26 in figure 5 before performing steps 21 and 22.

The parameters K, L, Δti, Δt ₂ , Δt _{3 as} well as the other internal parameters, in particular the size of the search window, the step for searching for delays considered in the second step, the maximum number of journeys taken into account for RAKE recombination, the threshold for path selection must be chosen according to the characteristics of the propagation channel and the compromise between performance and digital complexity to be respected.

Without departing from the scope of the invention, the learning or reference sequence is transmitted sequentially or in parallel to that of the useful data.

Likewise, without leaving the method according to the invention, the channel is for example estimated several times by time slot, the step of estimating the complex amplitude being carried out using a number of learning symbols less than the total number of learning symbols available on a time slot. This method of implementing the method is particularly advantageous for very rapidly fluctuating channels. It is also possible to make interpolation, so as to have at any time an estimate of a channel (linear filtering from several estimated values of the channel).

The method comprises for example a step of deciding the demodulated bits, these decided bits being able to be used to improve the estimation of the channel. The learning symbols can be consecutive or be distant in time. For example, other control symbols can be inserted between the learning symbols.

Example of control sequence

-: learning symbols: other type of control symbol (type 1)

===: other type of control symbol (type 2)

The estimation of the channel is obtained for example by simple correlation of the pilot symbols received with the spread learning sequence, which implicitly supposes that the spread learning sequence is white, its auto-correlation function being a Dirac. It is also possible to take account of the non-ideal nature of the training sequence by correlating the pilot symbols received with a bleached version of the training sequence.

According to another variant implementation of the method, the paths selected during the third step are not separated by an integer of chip duration.

ANNEX

List of key parameters of the invention and of the variables used in the algorithmic description.

Table 1: Key parameters of the algorithm

0) ^: Maximum value of the temporal spread of the radio mobile channel in an urban environment.

(2): The temporal resolution during the search for the paths is equal to T _c / (Nb_ech_chip / Sous_ech_retard) where T _c is the duration of the chip. The possible values of the temporal resolution of search for the paths are T _c 12, T _c 14 and T _c .

( ³ ) ^{: This value} represents the number of delay times considered in the search window. She wants : SizeJ ^" en_rech = 2 * (l / T _c ) * EtalJemp_max * (Nb_ech_chip / Sous_ech_retard) (656 = 2 * 4.096.10 ⁶ * 20.10 ⁶ * 4/1)

(4): This value means that any propagation path whose average power is greater than half of the maximum average power, is selected.

(5): This figure is expressed in number of time slots (see paragraph 2.2.1).

(6): This figure is expressed in a unit equal to K times the duration of the time slot (thus, the delay times are updated every K * L slots).

Table 2: Variables used in the algorithmic description

Claims

1 - Method for estimating a propagation channel, between a data transmission source and a receiver, using at least one training sequence comprising symbols known as “pilot symbols” and known to the receiver, to decode a data signal (digital signal), the signals being transmitted in successive frames, each frame being partitioned into a determined number of time slots characterized in that it comprises at least the following steps: a) choose a search window, estimate the complex amplitude and energy of each preponderant path k located in the search window for the duration of each of the time slots and demodulating the data bits (useful bits), b) updating the energy of each possible path of the search window every K slots, c) select all K ^* L slots, the preponderant energy paths k.

2 - Method according to claim 1 characterized in that the signal is a signal spread according to a given spreading code and in that it comprises at least: for step b)

• the determination of the energy or average power (Puiss noyjrajet (k)) for each path k of the search window by correlation between the pilot symbols received and the control spreading sequence, the average (Correl_pilot_moy (k)) said correlations on all the symbols of the learning sequence, the calculation of the correlation module and the updating of the average energy of the path (Puissjnoyjrajet (k)) by a calculation of the instantaneous average power type equal to the sum of the instantaneous power of said paths and of a term corresponding to the product of the forget factor and the average power,

• the determination of the complex amplitude (Channel_estimated p)) of said channel p by correlation between the pilot symbols and the spreading sequence. for step c)

The selection of propagation paths k by comparing the average energy of a path (Puissjmoyjrajet (k)) to a fixed threshold value (Puiss_moy_max), such as the maximum power of all the paths, • the determination of the complex amplitude of the propagation channel p (estimated channel (p)).

3- Method according to one of claims 1 or 2 characterized in that it comprises after the step of selecting the paths, a step of determining the best sampling comb (lnd_meilleur_peigne) by the argument of the total cumulative power maximum on all journeys exceeding the threshold value for all journeys separated by an integer number of chips and by determining the maximum cumulative power.

4 - Method according to one of the preceding claims characterized in that it comprises in step c) a step of comparing the number of paths (Nb-paths) selected with a maximum value and / or updating the average power of the paths (Puissjnoyjrajet) and of the index of the center of the search window (lnd_centre enetre) and the vector of the delays of the indices of the paths.

5 - Method according to one of claims 1 to 4 characterized in that

• K is fixed is a multiple of the duration of the time slot, and / or

• K ^* L is determined from the speed of variation of the paths over time.

6 - Method according to one of the preceding claims characterized in that it comprises a step of determining decided bits performed after the demodulation of the data bits or useful bits.

7 - Method according to one of the preceding claims characterized in that the learning sequence is transmitted sequentially or in parallel to that of the useful data.

8 - Device for estimating the parameters of a digital signal propagation channel between a transmission means and a reception means in order to demodulate a data signal (digital signal), the signals being transmitted by successive frames, each frame being partitioned into a determined number of time slots the useful data received, characterized in that it comprises a computer or device suitable for implementing the method according to one of claims 1 to 7.

9 - Device according to claim 8 characterized in that said transmitting means comprises a means for spreading the digital signal and said receiving means comprises a RAKE type receiver.

10 - Application of the method according to one of claims 1 to 7 or of the device according to one of claims 8 and 9 for demodulating a digital spread band signal.

1 1 - Application of the method according to one of claims 1 to 7 or of the device according to one of claims 8 and 9 for demodulating a signal of CDMA type, or of UMTS type.